Plasmon excitations of single-wall carbon nanotubes S. Dmitrović,* T. Vuković, B. Nikolić, M. Damnjanović, and I. Milošević Faculty of Physics, University of Belgrade, P.O. Box 368, Belgrade 11001, Serbia Received 22 January 2008; revised manuscript received 2 May 2008; published 11 June 2008 Plasmon excitations in isolated single-wall carbon nanotubes are singled out in the optical spectra, and analyzed within density-functional tight-binding method and random-phase approximation. Full symmetry considerations, implemented in both approaches, stressed out that only helical quantum numbers are conserved in processes involving momentum transfer, i.e., only these quantum numbers can be unambiguously attributed to the plasmons. Energy of plasmon is about 5 eV and slightly decreases with diameter with no observable influence of chirality. Energy of + plasmon is between 20 and 21 eV, and is insensitive to nanotube geometry. Dispersion of the plasmon is mainly linear. Slope of dispersion curve increases with tube’s translational period. DOI: 10.1103/PhysRevB.77.245415 PACS numbers: 61.46.Fg, 78.67.Ch, 71.45.Gm, 73.20.Mf I. INTRODUCTION Being one of the most important nanotechnological mate- rials, carbon nanotubes are examined in many details. 1 Still, there is a lack of complete insight to the collective electronic excitations plasmons, as the predictions in the existing studies differ due to the theoretical models applied. Experimentally, electron energy-loss spectroscopy EELSis the main tool for plasmon investigation in crystals. For single-wall carbon nanotubes SWCNTs, measurements in both aloof and penetrating modes have been performed. 25 Reported plasmons of isolated SWCNT Refs. 4 and 5are in the low-loss region up to 50 eV: a single surface plasmon at 15 eV and two volume-plasmon peaks around 5 and 23 eV. The high energy plasmon is associated with collective exci- tations of all valence electrons + plasmonwhile the lowest one comes solely from valence electrons plas- mon. One of the earliest measurements on SWCNT bundles, 2 with tubes of 1.0–1.3 nm in diameter, stated plasmon at energy of 5.8 eV and also the + plasmon at 20.6 eV. A single study 3 of momentum-dependent EELS for parallel polarization exists in the literature. These measure- ments are performed for bulk samples of purified SWCNTs, with mean diameter of 1.4 nm, showing the nondispersive excitations with energies below 4 eV and two dispersive ex- citations, one at 5.2 eV and another at 21.5 eV, for momen- tum transfer along the tubes’ axis. Prominent peak at 4.5 eV, independent of the tube’s diameter, has also been obtained in polarized optical-absorption measurements 6 on vertically aligned SWCNTs for parallel polarization stimulating further discussion about assignation of this peak to plasmon. When it comes to the reported theoretical studies, only a few of them goes beyond quasiclassical hydrodynamical model. However, quantum-mechanical treatment is necessary to encounter the effects of tube’s chirality, small wall thick- ness, etc. Zone folding of the graphene tight-binding TB electron bands had been used to calculate 7 dielectric matrix within the random-phase approximation RPAmainly for tubes with large diameter. Using a similar approach, but in- cluding local-field effects, 8 Perez and Que in Ref. 8 con- cluded that for chiral SWCNTs, the collective electronic modes are dispersive while for the achiral tubes modes are nondispersive. All of these studies are related solely to plasmon. Only a single study 9 of both experimentally re- ported volume plasmons is based on DFT calculations of both optical and energy-loss spectra of isolated and aligned SWCNTs with small diameters. These results are in excellent agreement with measurements, showing also that parallel to the tube axis, exchange-correlation and local-field effects were dumped. The aim of this paper is to thoroughly describe volume plasmons behavior of individual SWCNTs of an arbitrary geometry. Starting with full symmetry based and high- precision density-functional tight-binding DFTBelectronic Bloch states, we use two different approaches, 10,11 optical and RPA, as explained in Sec. II. The main results, i.e., and + plasmon energies and their dispersion properties, and also comparison with the interband transitions are pre- sented in Sec. III. II. METHOD Two different models accounted here are optical calcula- tions of the transversal dielectric function with zero trans- ferred momentum and RPA calculations of the longitudinal dielectric function needed for EELS calculations. Electronic energies and states are obtained by the full symmetry imple- mented DFTB method code POLSYM 12 . Full symmetry groups of SWCNTs Ref. 13are line groups L C = Lq p 22a for chiral tubes and L ZA = L2n n / mcmafor achiral ones; here, helicity p, and the orders q and n = GCDn 1 , n 2 = GCDq , pof the rotational axes of the isogonal group and of the line group are uniquely determined by the chiral indi- ces n 1 and n 2 , while the translational period a is found by relaxation. This symmetry implies that the Bloch eigenfunc- tions |kmand the corresponding eigenenergies k,m are assigned by the following quantum numbers: quasimomen- tum k -/ a , / a, z component of the angular momen- tum m =-q / 2,-q / 2+1,..., q / 2, and parities U , v and h invoked by U axis and, for achiral tubes only, horizontal and vertical mirror planestaking the values +1 for even and -1 for odd states while conventionally 0, otherwise. These quantum numbers are employed to get severe selection rules, significantly reducing the number of relevant matrix ele- PHYSICAL REVIEW B 77, 245415 2008 1098-0121/2008/7724/2454156©2008 The American Physical Society 245415-1